Just to add some elements to the discussion. Here's what I noticed on my Drake L-7, so with 2 x 3-500Z. The -9V bias on the grid is generated directly on the grid; the 9V zener is not in the cathode circuit.

?

?

Note that the grid current is lower when the plate voltage increases.

?

In France, we're limited to 500W, but I wonder how in the USA you can reach 1 kW without exceeding the maximum Eimac values.